Antistatic
Ask
Trade
Surface
Sign in
Auto
Light
Dark
← Internal deployments
Internal deployments
False information
All labs
OpenAI
Anthropic
Slice by
What goes wrong
Where it runs
Copies
Reasoning
Monitor reach
Pace
Fallback to a trusted model
▾
Sabotaging safety work (30%)
Running without over… (20%)
Gaming its training (20%)
Copying its wei… (15%)
Escaping its sa… (15%)
Sabotaging safety work · Training runs: 2% of surface, 0% covered (Authority to pause 4%, Monitoring of training runs 3%)
Coding agents
11% · 5%
Sabotaging safety work · Coding agents: 11% of surface, 5% covered (Alignment checks before internal use 11%, Probes on model internals 7%)
Research jobs
14% · 5%
Sabotaging safety work · Research jobs: 14% of surface, 5% covered (Alignment checks before internal use 11%, Probes on model internals 7%)
Evaluations
Sabotaging safety work · Evaluations: 5% of surface, 2% covered (Sabotage and sandbagging checks 5%, Authority to pause 4%)
Training runs
Running without oversight · Training runs: 3% of surface, 0% covered (Sandboxes and egress limits 21%, Sandbox-escape detection 13%)
Coding agents
10% · 3%
Running without oversight · Coding agents: 10% of surface, 3% covered (Sandboxes and egress limits 13%, Review of risky agent actions 12%)
Research jobs
6% · 3%
Running without oversight · Research jobs: 6% of surface, 3% covered (Sandboxes and egress limits 17%, Alignment checks before internal use 11%)
Running without oversight · Evaluations: 1% of surface, 1% covered (Sandboxes and egress limits 21%, Sandbox-escape detection 13%)
Training runs
16% · 0%
Gaming its training · Training runs: 16% of surface, 0% covered (Monitoring of training runs 12%, Authority to pause 4%)
Evaluations
4% · 0%
Gaming its training · Evaluations: 4% of surface, 0% covered (Monitoring of training runs 5%, Authority to pause 4%)
Training runs
5% · 0%
Copying its weights out · Training runs: 5% of surface, 0% covered (Sandboxes and egress limits 42%, Sandbox-escape detection 22%)
Coding agents
5% · 4%
Copying its weights out · Coding agents: 5% of surface, 4% covered (Sandboxes and egress limits 25%, Review of risky agent actions 14%)
Research jobs
3% · 4%
Copying its weights out · Research jobs: 3% of surface, 4% covered (Sandboxes and egress limits 34%, Sandbox-escape detection 13%)
Evaluations
Copying its weights out · Evaluations: 2% of surface, 1% covered (Sandboxes and egress limits 42%, Sandbox-escape detection 22%)
Training runs
5% · 0%
Escaping its sandbox · Training runs: 5% of surface, 0% covered (Sandboxes and egress limits 42%, Sandbox-escape detection 31%)
Coding agents
3% · 3%
Escaping its sandbox · Coding agents: 3% of surface, 3% covered (Sandboxes and egress limits 25%, Sandbox-escape detection 16%)
Escaping its sandbox · Research jobs: 2% of surface, 3% covered (Sandboxes and egress limits 34%, Sandbox-escape detection 19%)
Evaluations
5% · 1%
Escaping its sandbox · Evaluations: 5% of surface, 1% covered (Sandboxes and egress limits 42%, Sandbox-escape detection 31%)
Covered
46%
2+ layers
15%
Most exposed:
Escaping its sandbox · Training runs
(5% of surface · 0% covered)